Current text-to-avatar methods often rely on implicit representations (e.g., NeRF, SDF, and DMTet), leading to 3D content that artists cannot easily edit and animate in graphics software. This paper introduces a novel framework for generating stylized head avatars from text guidance, which leverages locally learnable mesh deformation and 2D diffusion priors to achieve high-quality digital assets for attribute-preserving manipulation. Given a template mesh, our method represents mesh deformation with perface Jacobians and adaptively modulates local deformation using a learnable vector field. This vector field enables anisotropic scaling while preserving the rotation of vertices, which can better express identity and geometric details. We also employ landmark- and contour-based regularization terms to balance the expressiveness and plausibility of generated head avatars from multiple views without relying on any specific shape prior. Our framework can generate realistic shapes and textures that can be further edited via text, while supporting seamless editing using the preserved attributes from the template mesh, such as 3DMM parameters, blendshapes, and UV coordinates. Extensive experiments demonstrate that our framework can generate diverse and expressive head avatars with high-quality meshes that artists can easily manipulate in 3D graphics software, facilitating downstream applications such as efficient asset creation and animation with preserved attributes.

Abstract Graph clustering is a fundamental task in unsupervised learning. Most studies use graph neural networks to learn node embeddings and then adopt traditional clustering methods. However, they usually face the following problems: (1) noise and sparsity in original graph structures and attributes hinder clustering. (2) When the intrinsic structure of the data is complex, the linear properties of Euclidean space may not sufficiently capture the features between different semantics. (3) Node representations in multiple views fail to connect or capture the global clustering structure effectively. To address these issues, we propose Multi-Curvature-based Adaptive Graph Fusion for Clustering. Firstly, we design a shared-neighbor filtering module to enhance the adjacency matrix with second-order neighbor proximity and denoise attribute features using a graph Laplace filter. Secondly, we employ graph convolutional networks and multivariate curvature techniques to generate feature views and geometric attribute views in Euclidean and Riemannian spaces, creating heterogeneous node embeddings that better capture data complexity. Finally, an attention mechanism and learnable view sampling vectors dynamically adjust view weights to capture semantic information from different perspectives. Extensive experimental results on six benchmark datasets validate the effectiveness of our model.

Remote collaborative design has become increasingly popular, but current design tools often overlook the importance of contextual communication during synchronized design activities, which is critical for understanding the rationale and decisions behind design choices. In this paper, we introduce DesignMemo, a proof-of-concept system that integrates the verbal context of remote discussions into visual design history tracking. The system automatically labels the visual elements with an annotation, which is linked to a certain transcript of the meeting, so that the user can easily recall the context of the visual design by clicking the element. The system also integrates an LLM agent for annotation-oriented summarization based on global context tracking, so users can quickly follow the rationale of the design without reading the lengthy transcript. Our user study with 24 participants suggests that the ability to track communication context makes the iterative design process smoother and more efficient.

In recent years, various head-worn virtual reality (VR) techniques have emerged to enhance object selection for occluded or distant targets. However, many approaches focus solely on ray-casting inputs, restricting their use with other input methods, such as bare hands. Additionally, some techniques speed up selection by changing the user's perspective or modifying the scene context, which may complicate interactions when users plan to resume or manipulate the scene afterward. To address these challenges, we present FocalSelect, a heuristic selection technique that builds 3D disambiguation through head-hand coordination and scoring-based functions. Our interaction design adheres to the principle that the intended selection range is a small sector of the headset's viewing frustum, allowing optimal targets to be identified within this scope. We also introduce a density-aware adjustable occlusion plane for effective depth culling of rendered objects. Two experiments are conducted to assess the adaptability of FocalSelect across different input modalities and its performance against five selection techniques. The results indicate that FocalSelect enhances selection experiences in occluded and remote scenarios while preserving the spatial context among objects. This preservation helps maintain users' understanding of the original scene and facilitates further manipulation. We also explore potential applications and enhancements to demonstrate more practical implementations of FocalSelect.

We introduce AniCraft, a mixed reality system for prototyping 3D character animation using physical proxies crafted from everyday objects. Unlike existing methods that require specialized equipment to support the use of physical proxies, AniCraft only requires affordable markers, webcams, and daily accessible objects and materials. AniCraft allows creators to prototype character animations through three key stages: selection of virtual characters, fabrication of physical proxies, and manipulation of these proxies to animate the characters. This authoring workflow is underpinned by diverse physical proxies, manipulation types, and mapping strategies, which ease the process of posing virtual characters and mapping user interactions with physical proxies to animated movements of virtual characters. We provide a range of cases and potential applications to demonstrate how diverse physical proxies can inspire user creativity. User experiments show that our system can outperform traditional animation methods for rapid prototyping. Furthermore, we provide insights into the benefits and usage patterns of different materials, which lead to design implications for future research.